This project was created for the final project of MIT's Hands-on Deep Learning class. It is a Smart Home AI Agent that leverages:
β
OpenAI API for energy price analysis & automation
β
ComEd & Bayou APIs for real-time & historical electricity data
β
Neural Networks (DNN) for future energy demand prediction
β Gradio Web UI for easy user interaction
- Fetch real-time energy prices from ComEd API
- Retrieve historical energy bills from Bayou API
- Predicts energy demand using a trained Neural Network (DNN)
- Automates HVAC settings based on AI-powered forecasts
- Users can ask for energy pricing, HVAC adjustments, or home occupancy
- Results are displayed in a simple web-based chatbot
Smart_Home_Agent_8.ipynb has neural network integration
Manually install:
pip install transformers gradio openai requests torch tensorflow keras PILOPENAI_API_KEY = "your-api-key"
COMED_API_URL = "https://hourlypricing.comed.com/api?type=5minutefeed&format=json"
BAYOU_API_KEY = "your-bayou-api-key"python gradio_ui.pyThen open http://localhost:7860 in your browser.
User: "What is the weather like in Chicago?"
Bot: "The current high temperature is 78Β°F."
User: "What is the current electricity price in Illinois?"
Bot: "The latest ComEd price is 14.2 cents/kWh."
User: "How much electricity will I use over the next 6 hours?"
Bot: "Based on your past usage, your predicted energy consumption for the next 6 hours is 1.8 kWh."
User: "Optimize HVAC settings based on energy prices."
Bot: "Energy price is high! Adjusting thermostat for efficiency."
User: "Optimize HVAC settings based on predicted demand."
Bot: "Energy usage is expected to be high. Adjusting thermostat to 70Β°F for efficiency."
User: "What is the current electricity price in Illinois?"
Bot: "The latest ComEd price is 14.2 cents/kWh."
β
Integrate with Smart Devices (e.g., Nest, Tesla Powerwall, Home Assistant API)
β
Enable Real-Time Image Processing from IP Cameras
β
Enhance AI Automation with Reinforcement Learning
β
Vision Transformers (ViT) for home occupancy detection
- Perform real-time home occupancy detection using ViT
- Uses Google's ViT model to analyze images
- Determines if a home is occupied or empty
- Can be expanded for security automation
Jennifer Turliuk, Siddharth Chilukuri, Dominic Sudnik, Kike Vera
This project is MIT Licensed β Feel free to modify & use. π
π Built for a Smarter, Energy-Efficient Home! π